Adam Li

To confirm this is not an isolated issue with comight right? Or so far it is?

The general MVN approach maybe can be done as Jovo suggested (w/ some open questions): X_i | Y ~ MVN, where for CoMIGHT, we generate two such instances that are...

Structuring the covariance in blocks as such and then using $Y \in [1, 2]$ to select the corresponding multivariate normal should allow us to: 1. arbitrarily apply feature-wise transformations for...

I tried again by removing some unnecessary for loops in the code and get this: ``` ============================== Iteration 020 of 025 ============================== n_samples: 125 and n_features: 50000 Unsupervised RF Speed:...

I suppose training 2x deeper trees should slow you down, but I'm surprised it's by an order of magnitude. I guess I would think in log terms since it's a...

My suspicion is that there is still a lot from moving the columns that are constant to the front of the array. This is what sklearn does in `main` and...

Cross-ref: #164

I think you are looking for the projection matrix and want to expose a Python API to access it? There is already `tree.tree_.get_projection_matrix()`. However, there is no Python API. If...

Yep seems right

Oh I guess the permuted forest technically gives that(?), but I was assuming you wanted like M forests each with a slightly different feature_importances map constructed from a different collection...